Regulation of immune system is of paramount importance to prevent immune attacks against self-components. Mice deficient in the interleukin (IL)-2/IL-15 receptor β chain, CD122, are model animals of such immune attacks and characteristically have a high number of abnormally activated T cells. Here, we show that the transfer of CD8+CD122+ cells into CD122-deficient neonates totally prevented the development of abnormal T cells. Furthermore, recombination activating gene–2−/− mice that received wild-type mice–derived CD8+CD122− cells died within 10 wk after cell transfer, indicating that normal CD8+CD122− cells become dangerously activated T cells in the absence of CD8+CD122+ T cells. CD8+CD122+ cells could control activated CD8+ or CD4+ T cells both in vivo and in vitro. Our results indicate that the CD8+CD122+ population includes naturally occurring CD8+ regulatory T cells that control potentially dangerous T cells.
Immune reaction against virulent foreign objects could be harmful for self-organisms. Therefore, a strict regulatory mechanism must exist in the immune system (1–6). Regulatory T cells are key subpopulations of T cells that control immune activity. CD4+CD25+ (IL-2 receptor α chain) regulatory T cells are a well-documented subpopulation of T cells (1, 2, 7–9). The importance of CD4+CD25+ regulatory T cells has been confirmed in many types of autoimmune diseases. However, other cells, especially those of CD8+ T cell lineage, have also been suggested as regulatory T cells (5, 10, 11).
CD122 (IL-2/IL-15 receptor β chain)-deficient mice exhibit severe hyperimmunity (12) with augmented granulopoiesis and suppressed erythropoiesis. Severe anemia progresses, and the mice die within 3–4 mo after birth. We found that expansion of abnormally activated T cells is the underlying cause of all phenotypes (13). The mechanism of generation of activated memory-type T cells in CD122-deficient mice is controversial. In the initial investigations by Abbas and others (14–16), the defect of programmed cell death (apoptosis) in IL-2 or IL-2Rα–deficient T cells could be applied to the mechanism for survival of activated CD122-deficient T cells, and such a possibility has not been excluded. Based on our previous experimental results, we proposed that the lack of certain regulatory cells was responsible for the abnormal phenotypes and that a subpopulation of CD8+ T cells could be involved in the control of CD122-deficient abnormal T cells (11). However, in our previous paper, direct evidence of the responsible subpopulation in CD8+ cells was not provided, and it was still unclear whether the lack of regulatory system is sufficient to cause T cell abnormalities in CD122-deficient mice, or whether intrinsic defect of CD122 molecules in T cells is required.
Here, we report a novel observation that the transfer of highly purified CD8+CD122+ T cells from normal mice into CD122-deficient neonates resulted in normalization of all the phenotypes in these mice. This result indicates that the CD8+CD122+ T cell population contains novel regulatory T cells that effectively control CD122-deficient T cells. In addition, we show that RAG-2−/− mice that received wild-type mice–derived CD8+CD122− cells die within 10 wk after cell transfer, indicating that normal CD8+CD122− cells without intrinsic defect of CD122 molecules become dangerously activated T cells in the absence of CD8+CD122+ T cells, which control not only CD122-deficient T cells but also wild-type T cells.
Materials And Methods
CD122-deficient mice were originally established by Suzuki et al. (12) and maintained in our animal facility. Breeding pairs of the C57BL/6CD45.1/CD45.1 congenic strain were obtained from Taconic and maintained in our animal facility. RAG-2–deficient mice with C57BL/6 genetic background were obtained from The Central Institute for Experimental Animals with permission from the founder F.W. Alt (Harvard Medical School) and maintained in our animal facility. OT-I transgenic mice bearing H-2Kb-restricted OVA-specific TCR (17) were gifts from T. Hirano and M. Murakami (Osaka University, Suita, Japan) The experimental protocol was approved by the Ethics Review Committee for Animal Experimentation of Nagoya University Graduate School of Medicine.
T Cell Clones.
FITC-conjugated anti–mouse CD8α (clone 53-6.7), PE-, or allophycocyanin-conjugated anti–mouse CD4 (clone GK1.5); biotin-conjugated anti–mouse CD122 (clone 5H4); FITC-conjugated anti–mouse CD25 (clone PC61.5); FITC- or biotin-conjugated anti–mouse CD45.1 (clone A20); and biotin-conjugated anti–mouse TER-119 (clone TER-119) antibodies were purchased from eBioscience. FITC-conjugated anti–mouse CD69 (clone H1.2F3), FITC-conjugated anti–mouse CD44 (clone IM7), PE-conjugated anti–mouse CD62L (clone MEL-14), FITC-conjugated anti–mouse Gr-1 (clone RB6-8C5), and PE-conjugated anti–mouse VLA-4 (clone 9C10) antibodies were purchased from BD Biosciences. Biotin-conjugated antibodies were visualized by streptavidin-PE-Cy5 (eBioscience).
Flow Cytometry and Cell Sorting.
Analytical flow cytometry was performed using FACS Calibur flow cytometer (BD Biosciences). Preparative cell sorting was performed using FACS Vantage cell sorter (BD Biosciences).
Adoptive T Cell Transfer to Neonates.
T cells were hypodermically transferred into 2–3-d-old neonates offspring of CD122-heterozygous mutant parents. After the neonates grew beyond 3 wk, their genotype for CD122 gene was analyzed by PCR from tail DNA (11).
Adoptive T Cell Transfer to RAG-2−/− Mice.
T cells were intravenously transferred into 8–10-wk-old RAG-2–deficient mice that had received a sublethal dose (5.5 Gy) of irradiation.
Peripheral blood was taken from mice tail vein using heparin-coated capillary tubes. After one end of the capillary was sealed with clay, it was centrifuged, and the length of packed RBC per total blood was measured.
Intracellular Cytokine Staining.
Intracellular cytokine staining was performed using a Cytofix/Cytoperm kit (BD Biosciences) according to the protocol provided by the manufacturer. Cells were incubated with biotin-conjugated anti–IFN-γ or anti–IL-2 antibody and visualized with streptavidin-PE-Cy5.
Total cellular RNA was extracted by using RNA-bee reagent (Tel-Test Inc.). The total amount of RNA was reverse transcribed using Superscript II reverse-transcriptase and random hexamer primers (Invitrogen). IFN-γ, TNF-α, and hypoxanthine-guanine phosphoribosyltransferase (HPRT) mRNA levels were quantified by real-time PCR using the ABI/PRISM 7700 sequence detection system (Applied Biosystems). Analyses were performed using primers, an internal fluorescent TaqMan probe specific to IFN-γ, TNF-α, or HPRT, and the TaqMan Universal PCR Master Mix (Applied Biosystems). Premixture of primers and TaqMan probe of IFN-γ and TNF-α were optimized for real-time PCR system by the manufacturers. Standard curves of cDNA from total CD8+ T cells and CD8+CD122− T cells were used to calibrate the threshold cycle to relative quantities of HPRT, TNF-α, and IFN-γ cDNA.
Age-related Distribution of CD8+CD122+ Cells in Wild-Type Mice.
In our previous paper, we found that the lack of regulatory T cells is the cause of abnormal phenotype in CD122-deficient mice (11). In the experiment of transfer of a mixture of wild-type T cells and CD122-deficient T cells, wild-type CD8+ T cells more effectively suppressed the abnormal activation of CD122-deficient T cells than wild-type CD4+ T cells (11). Based on this observation, we hypothesized that the regulatory T cells that control CD122− cells are in the CD8+ T cell population. Spleens (Fig. 1 a) and lymph nodes (not depicted) contain CD122+ cells in the CD8+ population. Because the expression level of CD122 fluctuates continuously among T cells both in CD8+ and CD4+, it may be difficult to distinguish CD122+ from CD122−. However, a population in CD8+ cells expresses obviously high levels of CD122, the same expression level of which is not observed in CD4+ T cells (Fig. 1 a). Therefore, though some CD4+ cells are actually CD122low+, we call the cells forming a population with a high CD122 expression level CD122+ cells. The percentage of CD122+ cells in the CD8+ population varied between 10 and 50% depending on mouse age (Fig. 1 b). Generally, the percentage of CD122+ cells in CD8+ population was high in very young mice, decreased to the lowest level of ∼10% at 7–10 wk of age, and increased in older mice. In contrast with the considerable proportion of CD122+ cells in the CD8+ population, CD122+ cells were rare in CD4+ population (Fig. 1 a, top) although CD4+ cells contain CD122low+ cells, many of which are CD25+ (not depicted). This is in contrast with the fact that CD25+ cells exist in the CD4+ population but not in the CD8+ population (Fig. 1 a, bottom).
Isolation of CD8+CD122+ Cells and Their Transfer to CD122-deficient Neonates.
CD8+CD122+, CD8+CD122−, and CD4+CD25+ cells were isolated from spleens and lymph nodes of 6-wk-old CD45.1 congeneic C57BL/6 mice by cell sorting. Fig. 2 a demonstrates that the purity of specific cell population after cell sorting was >98%. Next, 5 × 104 of these highly purified cells were subcutaneously transferred into CD122-deficient neonates, and the cells obtained from spleens or lymph nodes were analyzed 7 wk later. No donor-type CD4+ T cells were observed at 7 wk after transfer of CD8+ T cells, indicating that phenotypic changes, if any, should not be due to contaminated CD4+ T cells (Fig. 2 b). Donor-type CD8+ T cells were also very few in CD4+ T cell–transferred mice (Fig. 2 b). Analysis of the absolute numbers of each T cell population, derived either from host CD122-deficient mice or transferred wild-type donor cells, showed that transferred donor-type cells, especially CD8+ cells, expanded >20 times (from 5 × 104 to >106) during the 7 wk (Fig. 2 c). Furthermore, analysis of CD122-deficient neonates after transfer of 104 CD8+CD122+ cells derived from 2-wk-old or 2-yr-old mice showed that the resultant donor-type T cells 7 wk after transfer were mostly CD8+ cells and had expanded >50 times (from 104 to >106) during the 7 wk (Fig. 2 d).
CD8+CD122+ Cells Prevent Expansion of CD122-deficient Activated T Cells.
As shown in Fig. 3 (a and c, top middle), CD122-deficient mice without any treatment had a high percentage of activated memory-type T cells with markers of CD69+, CD44+, and CD62L−. In contrast, CD122-deficient neonatal mice that received 5 × 104 CD8+CD122+ cells showed no such increase of activated CD69+ T cells (Fig. 3 a, bottom left) and CD44+CD62L− activated memory-type T cells (Fig. 3 c, bottom left). Transfer of 5 × 104 CD8+CD122− T cells into CD122-deficient neonates did not prevent the development of activated CD69+ T cells (Fig. 3 a, bottom middle) and CD44+CD62L− memory-type T cells (Fig. 3 c, bottom middle). Transfer of 5 × 104 CD4+CD25+ T cells also did not alter the abnormal phenotype of CD69+ T cells (Fig. 3 a, bottom right) and CD44+CD62L− T cells (Fig. 3 c, bottom right) in CD122-deficient mice, although transfer of a larger number (2 × 105) of CD4+CD25+ T cells into neonates normalized all the phenotypes in CD122-deficient mice (not depicted). We also transferred 104 CD8+ CD122+ cells obtained from 2-wk-old or 2-yr-old mice into CD122-deficient neonates. This number of CD8+ CD122+ cells from 2-wk-old mice effectively normalized the shift to activated memory phenotype of T cells in CD122-deficient mice, whereas the same number of cells from 2-yr-old mice did not have such activity (Fig. 3, b and d). These results indicate that at least 104 of CD8+CD122+ cells in young mice are sufficient to perform regulatory activity when transferred at neonatal stage, but the same number of CD8+CD122+ cells in older mice does not contain a sufficient number of regulatory T cells. We analyzed CD4+ T cells and CD8+ T cells separately, and Fig. 3 shows representative results of CD122-deficient CD4+ T cells. Activation and memory status of CD122-deficient CD8+ T cells analyzed by expression of CD69, CD44, and CD62L were not different from those of CD4+ T cells (unpublished data).
CD8+CD122+ Cells Normalize Granulocytes, B Cells, and Erythrocytes in CD122-deficient Mice.
Examination of granulopoiesis showed that the number of Gr-1+ cells in the bone marrow was markedly increased in CD122-deficient mice (Fig. 4 a, top middle). This increase was normalized by transferring 5 × 104 CD8+CD122+ cells into neonates, but not by transferring 5 × 104 CD8+CD122− cells or 5 × 104 CD4+CD25+ T cells (Fig. 4 a, bottom). Changes in the number of Gr-1+ cells in the bone marrow matched those of leukocyte numbers in peripheral blood. Increased leukocyte number in peripheral blood of CD122-deficient mice was normalized by the transfer of 5 × 104 CD8+CD122+ cells into neonates but neither by the transfer of 5 × 104 CD8+CD122− cells nor by that of 5 × 104 CD4+CD25+ cells (Fig. 4 b). Transfer of 104 CD8+CD122+ cells from 2-wk-old mice normalized the number of Gr-1+ cells in the bone marrow (not depicted) and the leukocyte number in peripheral blood in CD122-deficient mice, but transfer of 104 CD8+CD122+ cells from 2-yr-old mice did not (Fig. 4 b). Severe anemia caused by a combination of autoimmune hemolysis and suppression of erythropoiesis is another striking feature of CD122-deficient mice (12). Examination of erythroid lineage cells by expression of TER-119 antigen (20), an erythroid specific marker, showed marked reduction in the number of TER-119+ cells in the bone marrow of CD122-deficient mice. This erythroid suppression was also rescued by the transfer of 5 × 104 CD8+CD122+ cells into neonates, whereas the transfer of 5 × 104 CD8+CD122− cells or 5 × 104 CD4+CD25+ cells was ineffective, resulting in the same decrease of TER-119+ cells as that in nontreated CD122-deficient mice (Fig. 4 c). Analysis of hematocrit in peripheral blood also confirmed the prevention of anemia by the transfer of CD8+CD122+ cells but not by that of other cell populations (Fig. 4 d). The number of B220+ cells, representing all developing B cells, was also normalized by the transfer of 5 × 104 CD8+CD122+ cells, but not by that of 5 × 104 CD8+CD122− cells or 5 × 104 CD4+CD25+ cells (Fig. 4 c). Again, transfer of 104 CD8+CD122+ cells from 2-wk-old mice rescued erythropoiesis in the bone marrow (not depicted) and normalized hematocrit in CD122-deficient mice, whereas transfer of 104 CD8+CD122+ cells from 2-yr-old mice did not (Fig. 4 d). As a conclusive result, CD122-deficient mice that received CD8+CD122+ cells at neonatal period grew normally and remained healthy. No improvement of general condition was observed in mice that received 5 × 104 CD8+CD122− or 5 × 104 CD4+CD25+ cells, and they all died before 3 mo of age. Transfer of a smaller number (104) of CD8+CD122+ cells from 2-wk-old mice at the neonatal stage effectively normalized all phenotypes in CD122-deficient mice, but that of the same number of CD8+CD122+ cells from 2-yr-old mice did not. 104 CD4+CD25+ cells obtained from 2-wk-old mice were also ineffective in normalizing CD122-deficient mice (unpublished data).
CD8+CD122+ Cells Control Both CD8+ and CD4+ Cells.
From the aforementioned experimental results, it is clear that CD8+CD122+ cells control CD122-deficient T cells. To elucidate whether CD8+CD122+ cells can control already-activated T cells, we transferred a mixture of CD8+CD122+ cells and already-activated T cells obtained from CD122-deficient mice into RAG-2−/− mice and analyzed the status of T cells 1 wk later. Activated memory-type cells (CD44+CD62L−) were dominant among CD8+ and CD4+ T cells from CD122-deficient mice (Fig. 5, a and c, before). After 1 wk of coexistence with CD8+CD122+ cells, naive-type cells (CD44−CD62L+) became dominant both in CD8+ and CD4+ cells (Fig. 5, a and c, +CD8+CD122+). In contrast, coexistence with CD8+CD122− cells did not alter the ratio between naive-type and activated memory-type cells (Fig. 5, a and c, +CD8+CD122−). The absolute number of naive-type cells increased, but that of activated-type cells rather decreased when CD8+ and CD4+ T cells coexisted with CD8+CD122+ cells (Fig. 5, b and d). CD122-deficient T cells that had been labeled with CFSE and coexisted with CD8+CD122+ cells for 7 d showed the same pattern of CFSE intensity as those that had coexisted with CD8+CD122− cells (Fig. 5 e), indicating that naive-type cells that appeared after coexistence with CD8+CD122+ cells were not the cells expanded from small number of naive-type cells or stem cells. These results (Fig. 5, b, d, and e) may better explain the phenomenon observed in Fig. 5 (a and c) as the phenotypic change of cells from activated type to naive type than by simple expansion of naive cells or deletion of activated cells. Although the only action to the CD122-deficient mice–derived T cells has been demonstrated eventually, these results suggest that CD8+CD122+ cells control already-activated T cells of both CD8+ and CD4+ types.
Next, we investigated the in vitro action of CD8+ CD122+ cells. Wild-type mice–derived CD8+CD122− cells cocultured with CD8+CD122+ cells showed lower expression of IFN-γ than those cultured alone (Fig. 6 a). Wild-type mice–derived CD4+CD25− cells cocultured with CD8+CD122+ cells also showed low expression of IFN-γ (Fig. 6 b) and IL-2 (not depicted), suggesting that CD8+CD122+ cells can potentially suppress cytokine expression in both CD8+ and CD4+ cells derived from wild-type mice. Time-course study of IFN-γ expression indicated that IFN-γ expression was induced in a time-dependent manner in CD8+CD122− cells cultured alone with plate-bound anti-CD3 and rIL-2, whereas that in CD8+CD122− cells cocultured with CD8+CD122+ cells was clearly suppressed through the culture during 2–5 d (Fig. 6 c). In this in vitro experiment, only purified CD8+ and CD4+ T cells were used and no APCs were added to the culture. When purified CD8+CD122− cells from OT-I transgenic mice that express OVA-specific TCR were stimulated with OVA peptide presented by MHC class I H-2Kb on syngeneic APCs for 4 d and cocultured with CD45.1-congenic CD8+CD122+ T cells during the latter 2 d of culture, CD8+CD122+ T cells obtained from C57BL/6-CD45.1 mice suppressed IFN-γ production from transgenic T cells (Fig. 6 d). In addition, CD8+CD122+ cells suppressed IL-2 production from Th1-type clones that were stimulated with antigen and OVA presented by MHC class II I-Ab on syngeneic APCs (Fig. 6 e). In these cases, using OVA peptide–responding T cells, CD8+CD122+ cells were not stimulated with anti-CD3 antibody, indicating that nonspecific activation was unnecessary for performing the regulatory action in CD8+CD122+ cells. We also examined the activity of CD8+CD122+ cells in suppressing proliferation of regulated cells. The results showed that CD8+CD122+ cells suppressed the proliferation of anti-CD3–stimulated CD8+ CD122− cells when these cells were cocultured without exogenous IL-2 (Fig. 6 f). This suppression of proliferation could be due to reduced production of IL-2 from regulated cells or to direct suppression effect on cell proliferation. The cytotoxic activity of CD8+CD122+ cells against anti-CD3–activated CD8+CD122− cells or CD4+CD25− in a culture condition, similar to that used for detecting suppression of cytokine production, was not detected (unpublished data), suggesting that cytotoxic activity against the regulated cell is not involved in the regulatory action of CD8+CD122+ cells.
Normal CD8+CD122− Cells without CD8+CD122+ Cells Cause Death of RAG-2−/− Mice.
As the aforementioned in vitro experiment suggested a regulatory activity of CD8+ CD122+ cells to normal T cells, we examined whether CD8+CD122+ cells control normal T cells and how normal T cells behave in the absence of CD8+CD122+ cells in vivo. The CD8+ cell population of normal 6-wk-old C57BL/6 mice consists of ∼10% CD122+ cells and 90% CD122− cells (Fig. 7 a). We collected CD8+CD122− cells and total CD8+ cells, each with a purity of >99%, by cell sorting and transferred 5 × 105 of these cells into lymphocyte-deficient RAG-2−/− host mice (Fig. 7 a). Surprisingly, all mice that received CD8+CD122− cells died within 10 wk after cell transfer, whereas all mice that received total CD8+ cells were alive and healthy for >20 wk after cell transfer (Fig. 7 b).
Normal CD8+CD122− Cells Become Highly Activated Cells in the Absence of CD8+CD122+ Cells.
We examined the mice at 7 wk after cell transfer into RAG-2−/− mice. All tissue of mice that had received CD8+CD122− cells appeared pale and anemic. Total cell numbers in the spleen and bone marrow of bilateral femurs of mice that received total CD8+ cells were not significantly different from those that received CD8+CD122− cells (spleen: 1.83 ± 0.08 × 107 vs. 1.64 ± 0.26 × 107; bone marrow: 5.01 ± 0.25 × 107 vs. 4.80 ± 0.28 × 107, respectively, n = 5 each, mean ± SD). We analyzed the status of transferred CD8+ cells. Although the numbers of CD8+ cells recovered from spleens were not different between the two groups (Fig. 8 a), the percentage of activated CD69+ cells was higher in CD8+CD122− cell transfer than in total CD8+ cell transfer (Fig. 8 b). We also transferred 5 × 105 CD8+CD122− cells mixed with 5 × 104 CD8+CD122+ cells or 5 × 103 CD8+CD122+ cells into RAG-2−/− mice. In the case of transfer of CD8+CD122− cells mixed with 1/10 the number (5 × 104) of CD8+CD122+ cells, the pattern of total CD8+ cell number and the percentage of CD69+ cells in the spleen were similar to those in total CD8+ cell transfer (Fig. 8, a and b). In contrast, the pattern of total CD8+ cell number and the percentage of CD69+ cells in spleen were similar to those in CD8+CD122− cell transfer in the case of transfer of CD8+CD122− cells mixed with 1/100 number (5 × 103) of CD8+CD122+ cells. The activated cells that had originally been CD122− became CD122+ (Fig. 8 c). In the bone marrow, the difference between CD8+CD122− cell transfer and total CD8+ cell transfer was more evident. The number of CD8+ T cells in the bone marrow of CD8+CD122− cell-transferred mice was ∼20 times greater than that of total CD8+ cell-transferred mice (Fig. 8 d). The CD8+ cells in the bone marrow of CD8+CD122− cell-transferred mice expressed high levels of the activation marker, CD69 (Fig. 8 e), which was higher than that of T cells in spleens of the same mice (Fig. 8, b and e). Again, the total CD8+ cell number and the expression pattern of CD69 in the bone marrow of mice that received 5 × 105 CD8+CD122− cells mixed with 5 × 104 CD8+CD122+ cells was similar to that in total CD8+ cell-transferred mice, whereas that in the bone marrow of mice that received 5 × 105 CD8+CD122− cells mixed with 5 × 103 CD8+CD122+ cells was similar to that in CD8+CD122− cell-transferred mice (Fig. 8, d and e). Real-time RT-PCR analysis showed higher expression levels of IFN-γ and TNF-α in T cells from CD8+CD122− cell-transferred mice than those in total CD8+ cell-transferred mice (Fig. 8 f). These features of T cells in CD8+ CD122− cell-transferred mice were similar to those in CD122-deficient mice, except that the absolute numbers of T cells in lymph nodes and spleens of the former mice were smaller than those of the latter mice.
Hematopoietic Disorders in CD8+CD122− Cell-transferred Mice.
The percentage of granulocytic Gr-1+ cells was increased to >90% in the bone marrow of CD8+CD122− cell-transferred mice compared with ∼60% in total CD8+ cell-transferred mice (Fig. 9 a). Alternately, cells in erythroid lineage that expressed TER-119 almost disappeared in CD8+CD122− cell transfer (Fig. 9 a). Most of the TER-119+ cells in the bone marrow were VLA4+ erythroblasts (Fig. 9 a, bottom). These bone marrow changes matched changes in the numbers of red and white blood cells in peripheral blood. The number of leukocytes was markedly increased (Fig. 9 b), and hematocrit was significantly reduced (Fig. 9 c) in CD8+CD122− cell-recipient mice, indicating progression of anemia in these mice. RAG-2−/− mice that had received 5 × 105 CD8+CD122− cells mixed with 5 × 104 CD8+CD122+ cells had a normal TER-119+/Gr-1+ cell ratio in the bone marrow (26% of TER-119+ cells and 62% Gr-1+cells; not depicted), normal leukocyte number in peripheral blood (Fig. 9 b), and normal hematocrit (Fig. 9 c). In contrast, RAG-2−/− mice that had received 5 × 105 CD8+CD122− cells mixed with 5 × 103 CD8+CD122+ cells had a TER-119+/Gr-1+ cell ratio skewed toward Gr-1+ cells (2% of TER-119+ cells and 91% of Gr-1+ cells; not depicted), increased leukocyte number in peripheral blood (Fig. 9 b), and low hematocrit (Fig. 9 c).
Regulatory T cells are classified into two groups; i.e., naturally occurring regulatory cells and regulatory cells induced during immune responses (5). A representative example of the former type is in the CD4+CD25+ population. In this paper, we clearly demonstrated that CD8+ cells contain regulatory T cells reminiscent of naturally occurring type, and that such regulatory cells are relatively rich in CD8+CD122+ population, but are rare in the CD8+CD122− population.
In recent studies by Malek et al. (21, 22), the defect of CD4+CD25+ regulatory T cells was proposed to be responsible for the increase of deregulated T cells in CD122-deficient mice. Based on the fact of the presence of very few CD4+CD25+ T cells in CD122-deficient mice, this idea might well explain the development of abnormalities in CD122-deficient mice. However, in the present paper, transfer of highly purified CD4+CD25+ cells failed to rescue the phenotypes in CD122-deficient mice (Figs. 3 and 4). These experimental results were incompatible with those of Malek et al. (21). In their work, CD122-deficient mice in which CD4+CD25+ T cell population had been transferred at neonatal stage were rescued from the development of abnormal phenotypes. There are several possibilities to explain the different results of the two papers. In our experiment, the number of effective CD4+CD25+ regulatory cells transferred by subcutaneous injection of 5 × 104 cells might not be sufficient to induce regulatory activity. Actually, transfer of a higher number (2 × 105) of CD4+CD25+ in our experiment was effective in normalizing all the phenotypes in CD122-deficient mice. Thus, the regulatory activity of CD4+CD25+ cells for CD122-deficient T cells was also confirmed in our paper. However, effective normalization of CD122-deficient mice by transfer of CD8+CD122+ cells but not by that of the same number of CD4+CD25+ cells under the same experimental condition (Figs. 3 and 4) allows us to make a definite conclusion that CD8+CD122+ cells contain stronger regulatory T cells that can control CD122-deficient T cells compared with CD4+CD25+ cells.
The CD8+CD122+ regulatory cells are exclusively important in the control of CD8+CD122− cells. CD4+ CD25+ regulatory cells do not seem to control normal CD8+CD122− cells similar to CD8+CD122+ cells because transfer of a mixture of CD4+CD25+ cells (105) and wild-type mice–derived CD8+CD122− cells (4 × 105) into RAG-2−/− mice resulted in the same lethal phenotype as that of simple CD8+CD122− cells (5 × 105; unpublished data). A mixture of one tenth the number (5 × 104 vs. 5 × 105) of CD8+CD122+ cells in CD8+CD122− cells prevented the development of lethal phenotype caused by deregulated activation of CD8+ T cells, whereas a mixture of one hundredth the number (5 × 103 vs. 5 × 105) of CD8+CD122+ cells in CD8+CD122− cells did not (Figs. 8 and 9), indicating that CD8+CD122+ cells ranging from one hundredth to one tenth of the number of CD8+ CD122− cells are necessary to maintain homeostasis in activation of CD8+ T cells. CD8+CD122+ T cells, along with CD8+CD122− cells, do not express Foxp3 (unpublished data), also suggesting that CD8+CD122+ cells are different cell lineage from CD4+CD25+ cells. Although the precise mechanism of regulation by CD8+CD122+ cells is not clear at present, the experimental results of in vitro cocultures (Fig. 6, a–c) suggest a direct interaction between the regulatory cells and regulated cells without intervention of APCs. This feature is different from that of CD4+CD25+ regulatory T cells, which require the interposition of APCs for regulatory action (23–25). Another exclusive feature of CD8+CD122+ cells may be that they possibly change the character of already-activated T cells, at least in case of CD122-deficient mice–derived T cells (Fig. 5, a–d).
Apart from the aforementioned difference between CD4+CD25+ regulatory T cells and CD8+CD122+ cells, it is quite an interesting correlation that CD25 (IL-2 receptor α chain) and CD122 (IL-2 receptor β chain) could be markers for regulatory T cells in CD4+ populations and CD8+ populations, respectively. IL-2 seems to be essential for maintaining already-developed CD4+CD25+ regulatory T cells (21). For the development of CD4+CD25+ cells, IL-2 receptor β chain is not essential because CD122-deficient mice that were cured by inoculating CD8+CD122+ regulatory T cells at neonate developed CD4+CD25+ cells that were not genetically affected (unpublished data). Alternately, IL-2 receptor β chain may be rather required in the development of CD8+CD122+ regulatory T cells in the thymus because the exogenous expression of CD122 only in the thymus is sufficient to normalize the phenotype of CD122-deficient mice (21). IL-2 does not seem to be indispensable for the development of CD8+CD122+ regulatory T cells because IL-2–deficient mice do possess CD8+ CD122+ cells (unpublished data) that could possibly be maintained by the compensatory action of other cytokines, including IL-15. The CD8+CD122+ regulatory T cells cannot be induced from CD8+CD122− cells even if they restored the expression of CD122 molecule. Actually, some CD8+CD122− cells transferred into CD122-deficient neonates (unpublished data) or into RAG-2−/− mice (Fig. 7 c) become CD122+, suggesting that expression of CD122 molecule is not sufficient to provide regulatory activity.
In the same way that CD25 (which is also expressed in activated T cells) is not an exclusive marker for CD4+ regulatory T cells, CD122 cannot be an ideal marker for CD8+ regulatory T cells. CD8+CD122+ populations have been considered memory-type T cells and their number actually increases in old mice. The recently identified subpopulations of memory T cells called “central memory T cells” have a cell surface marker of CD8+CD122+CD62L+ (26–32). Alternately, CD8+CD122+ cells, which also express CD62L, are relatively abundant in neonatal mice (>50% of total CD8+ in 1-wk-old mice; Fig. 1 b). The change in percentage of CD8+CD122+ cells in CD8+ cells, which show two age-related phases (Fig. 1 b), may suggest two different components in this population. As the number of memory T cells is increased in older mice, the increased ratio of CD8+CD122+ cells at old age may be due to the increase of central memory T cells (29, 30, 33). Instead, a higher rate of regulatory T cells may contribute to the high ratio of CD8+CD122+ cells in younger mice. Thus, the CD8+CD122+ population may include both the regulatory T cells and the central memory T cells. These results emphasize the need for markers that distinguish CD8+CD122+ regulatory T cells from the central memory T cells.
We noted extreme activation of CD8+ cells in RAG-2−/− mice that received CD8+CD122− cells. How such activated T cells are generated and which antigens they are responding to remain unclear at present. The profile of TCR Vβ usage of activated CD8+ T cells in CD8+CD122− cell-transferred mice was not significantly different from that of nonactivated T cells in total CD8+ cell-transferred mice (unpublished data), suggesting that the activated T cells were polyclonal cells responding to multiple antigens. Similar activation of T cells was observed in CD122-deficient mice, in which the activated T cells were also polyclonal (13), further supporting polyclonal activation of CD8+ T cells in CD8+CD122− cell-transferred mice. The activated T cells in CD8+CD122− cell-transferred mice, and possibly those in CD122-deficient mice as well, could be responding to multiple self-antigens, foreign antigens, or both. Whatever the antigens are, activated T cells may disappear in a usual condition with CD8+CD122+ regulatory cells but remain without such regulatory cells (Fig. 5, a–d).
In this paper, we demonstrated that the regulatory cells that control CD122− T cells are in the CD8+CD122+ population, but not in the CD8+CD122− population. We propose here that CD8+CD122+ T cells include a new type of regulatory T cells that effectively regulates other T cells. Normal CD8+CD122− T cells become dangerously activated cells that cause lethality unless the regulatory T cells in CD8+CD122+ population properly regulate them.
We thank H. Shiku for careful reading and critical comments on this paper; T. Kato and K. Kuribayashi for providing T cell clones; M. Murakami and T. Hirano for providing transgenic mice; M. Ito for histology; and K. Takeda, T. Shi, K. Watanabe, A.T. Endharti, and Y. Yi for experimental help.
This work was supported by grants from the Ministry of Science, Education, Sports and Culture of Japan.
The authors have no conflicting financial interests.
Abbreviation used in this paper: HPRT, hypoxanthine-guanine phosphoribosyltransferase.